5 IPv6 Attacks and Countermeasures v1.2

800.800.4239 | CDW.com/peoplewhogetit

IPV6 ATTACKS AND COUNTERMEASURES

CDW Advanced Technology Services

James Small, Principal Network/Security Consultant

2 2 CDW ADVANCED TECHNOLOGY SERVICES

PROBLEMATIC APPROACHES TO IPV6

Image Source: wordpress.com Image Source: thethingaboutflying.com

If I ignore it, nothing will happen

If I deploy it, Ill get DoSed!


VPN Bypass

Router Advertisement Spoofing/Flooding

DHCPv6 Spoofing

Remote Scanning/DoS Attack

Monitoring and Detection

Preventing Tunneling and Firewalling

Loss of NAT Security

Q&A throughout, I may postpone questions until the end depending on time

IPV6 PROBLEMS WELL PUT TO REST


VPN Bypass


DHCPv6 Spoofing





ROADMAP


Many organizations allow remote user VPN access to their networks

Often times access control and/or firewall policies are pushed to the client

However, these policies are typically IPv4 only

Many also disallow or restrict split tunneling the ability to send network traffic without going through the VPN session

One risk of allowing split-tunneling is that the remote user VPN client could be used as a bridgehead into the organizations network

Remote user connects to organization via VPN

Attacker compromises user system, goes from the Internet through the system into organization

VPN BYPASS


Belief: My VPN solution does not allow split tunneling all traffic is forced through the VPN:

Expected view of system traffic:

VPN BYPASS


Reality: All IPv4 traffic is forced over the VPN, IPv6 traffic completely bypasses it

If the system receives an IPv6 Router Advertisement it will immediately configure IPv6:

This may include a global address, a default route, and a new DNS server

This new IPv6 address, default route, and DNS server will be preferred over the IPv4 options (See RFC 6724/3484)

Do you see anything concerning about this full-tunnel VPN client traffic:

VPN BYPASS


Security challenges

Accidental VPN Bypass User has IPv6 at home or uses a dual stack network

Is preventing split-tunneling important?

Does the clients endpoint security protect against IPv6 attacks?

Malicious VPN Bypass Attacker injects Router Advertisement to configure IPv6 on users computer

With control of DNS and IPv6, the attacker can

- sniff all client traffic

- attempt Man-In-The-Middle attacks

- impersonate servers/systems and capture presented user credentials (e.g. NTLM)

- gain access into your organizations network

VPN BYPASS


Solution

Typically the vendors current VPN solution supports IPv6

For this particular case, the vendor has had a solution since early 2010

Test your solution!

Ideally a VPN solution allows:

Full support for either IPv4, IPv6 or both

Supports IPv6 over IPv4 and IPv4 over IPv6

Allows blocking/disabling either IPv4 or IPv6

Allows VPN bypass of either IPv4 or IPv6 (but only on purpose!)

Allows application of ACLs for either IPv4 or IPv6

Allows pushing firewall policy for either IPv4 or IPv6

VPN BYPASS - REMEDIATION

See Appendix for issues with disabling IPv6


VPN Bypass


DHCPv6 Spoofing





ROADMAP


Brief recap of the changes from IPv4 to IPv6

In particular, fragmentation changes are important

As you know, IPv6 eliminates header options:

Fixed length base header, fragmentation not handled here

IPV6 CHANGES QUICK REFRESHER


Review IPv6 header changes

All options now Extension Headers including Fragmentation

Extension headers/Upper Layer Protocols are not required to be in the first packet

If ULP not in first packet, stateless ACLs can be bypassed

IPV6 FRAGMENTATION


What if the IPv6 Upper Layer Protocol isnt in first packet?

Mitigation against surreptitious fragmentation

Stateful inspection/ACL

Effective, but not realistic for all access ports/points

Stateless ACL options

Deny undetermined-transport new option to block initial fragments without an Upper Layer Protocol

- Caution this also blocks OSPFv3, make sure to allow this if needed!

Deny fragments blocks non-initial fragments

IPV6 FRAGMENTATION CONTROLS

Initial Fragment

2nd Frgmt, ULP


By default, Windows Vista and newer, OS X, and Linux have IPv6 enabled

Many networks are only designed for IPv4 with no controls for IPv6

What happens when an IPv6 enabled system receives a router advertisement?

ROUTER ADVERTISEMENTS


Security challenges

Accidental RA

User with Windows Internet Connection Sharing service enabled (think BYOD or power user)

Someone connects a device configured for IPv6 routing to the network

Malicious RA

Attacker injects to attack network nodes as described in VPN Bypass section

Attacker uses to flood the network as Denial of Service (DoS) attack

ROGUE ROUTER ADVERTISEMENTS (RA)


Block RAs on unauthorized ports

RA Guard (If available) ipv6 nd raguard policy HOST

device-role host

!

vlan configuration 101

ipv6 nd raguard attach-policy HOST

ACL: ipv6 access-list HOST_PORT

remark Block RAs on Host Ports

deny icmp any any router-advertisement

permit ipv6 any any

!

interface GigabitEthernet0/1

description Host Port

ipv6 traffic-filter HOST_PORT in

ROGUE RA MITIGATION FIRST TRY


Does RA Guard or an IPv6 ACL work?

Yes for non-malicious RAs

- Test Windows 7 Workstation with Router on same VLAN

- Router connected to switchport with ACL or RA Guard on VLAN

- Router continuously generates RAs:

- Check Workstation No routable IPv6 address!



Does RA Guard or an IPv6 ACL work?

What about malicious RAs?

- Same Windows 7 Workstation with Linux Workstation on same VLAN

- Linux Workstation connected to switchport with ACL or RA Guard on VLAN

o RA generated by SI6 Networks IPv6 Toolkit (ra6)

- Check Workstation Uh oh



How were the ACLs and RA Guard evaded?

The fragmentation trick we showed earlier!

ACL Mitigation (Using previously shown options)

Use the undetermined-transport option (If available)

- I went to try this out on my Cisco 3k access switch:

- Then I discovered in the 2k/3k access switch configuration guide:

o The switch does not support matching on these keywords: flowlabel, routing header, and undetermined-transport.

ROGUE RA CONTROLS SECOND TRY, A


Mitigation against fragmented rogue RAs continued:

ACLs using the fragments option

Drawback is what to block a little more work

Is blocking packets to ff02::1 sufficient? Unfortunately no.

Sending RA to any multicast group the host is listening to or its link-local address activates IPv6 must block them all

- Windows: - Linux:

ROGUE RA CONTROLS SECOND TRY, B




Multicast/Link-local block candidates:

Most dangerous

- ff02::1 (all nodes on link)

- ff02::c (SSDP Windows)

- ff02::fb (MDNS OS X, Linux)

- ff02::1:3 (LLMNR Windows)

Harder to attack but possible

- ff02::1:ff00:0/104 (Solicited Node Multicast)

- fe80::/64 (all link-local addresses)

Unlikely, only configure if in use

- fe80::/10 (defined link-local only fe80::/64 should be used but some systems allow)

- ff02::/16 (all link-local multicast, also ff[137]2::/16)

ROGUE RA CONTROLS SECOND TRY, B




Reasonable ACL for most cases:

Of course, if your nodes listen on other IPv6 multicast groups you have to add those too

ROGUE RA MITIGATION SECOND TRY, B


While discussing this with Enno Rey he pointed out that actually the undetermined-transport option does work!

Documentation/Error messages bah!

A few options:

Apply the PACL to the port without the undetermined-transport ACE

After the PACL is applied then add the optionand it works!

Or:

Add an empty PACL to the port

Then create the ACL entries:

ROGUE RA MITIGATION THE END?


Caveats

If you want to apply it to other ports you have to remove the undetermined-transport option and add it back

Reboots do not seem to be a problem but there may be other quirks

May not be supported by Cisco TAC

ROGUE RA MITIGATION THE END?


One Denial of Service attack that gets repeated press is router advertisement flooding

A system connected to your LAN can flood RAs causing a DoS for many systems including:

100% CPU Utilization

Hanging/Crashing/Rebooting

But

Only works against systems on same LAN (L2 adjacent)

Typically requires high speed network with quality switch (wont work with something from Best Buy!)

Generally doesnt work over Wireless

Requires some work and only a DoS, cant exploit so better attacks available

RA FLOODING


Tools:

The Hackers Choice, thc-ipv6 suite (attack/fuzzing tools)

fake_router6, flood_router26

SI6 Networks IPv6 Toolkit (really meant for fuzzing/hardening)

ra6

Tested Attacks with:

Quad-core i7 high end laptop that generates 120,000 pps

Running Ubuntu 12.10

Cisco 3000 series gigabit switch (C/E/X-Series)

IP Base, 15.0(2)SE

RA Guard (as shown previously)

IPv6 ACLs (as shown previously)

RA FLOODING TEST SETUP


In a pristine lab environment:

Use fake_router6 and flood_router26

flood_router26 generates 17 prefixes and 17 routes per RA

Sends them as fast as possible to overwhelm host

RA FLOODING - OVERVIEW

(Lots of Prefix/Route Information options omitted)


Test 1 no defenses

Use fake_router6 and flood_router26, no options

Windows 7 (with KB2750841) unusable when flooded but recovers quickly when flood ends

Windows Vista becomes unusable, sometimes crashes

Windows 8 Crashes

RA FLOODING - RESULTS


Test 2 use RA Guard

Use fake_router6 and flood_router26

Blocks with no options but

Can bypass RA Guard with D (uses fragmentation tactics described earlier)

However:

- With a blazing fast laptop attacking I still couldnt crash Windows 8 or Vista and none of the systems were unresponsive

- But does trash IPv6 configuration all IPv6 addresses/routes are overwritten

Using SI6 Networks ra6

With carefully crafted fragmented packets it is still possible to crash Windows 8 and Vista, but hard and not consistent

7 is as before worst you can do is bog it down, but quickly recovers subsequent to attack



Test 3 use IPv6 ACLs described previously (undetermined-transport/fragments)

Use fake_router6, flood_router26, and ra6

Some fragments (initial or subsequent) get through but are harmless

No noticeable effect on the systems

Bottom Line - Make sure to Test your Equipment!

Older hardware/ASICs have limited or no support for these features make sure to read the documentation and validate the configuration!



VPN Bypass


DHCPv6 Spoofing





ROADMAP


By default, some operating systems such as Windows (Vista and newer) try to configure IPv6 via DHCPv6 (even without RA!)

The issues are similar as described for Router Advertisements

What happens when an IPv6 enabled system receives a DHCPv6 response?

It will configure an IPv6 address

It will configure a DNS server

It will configure a DNS search list

Note: It wont configure any routes or a default gateway these must come from RAs!

DHCPV6 SPOOFING


Security challenges similar to RAs

Accidental

Someone connects a device configured for DHCPv6 to the network

Malicious

Attacker responds to a DHCPv6 request with spoofed information

Notes:

- Not as easy as spoofed RAs attacker must respond to client requests with valid information making it easier to trace

- Not as dangerous as rogue RAs, primary threat is attacker gaining control of DNS

ROGUE DHCPV6 SERVER


Block DHCPv6 on unauthorized ports

DHCPv6 Guard ipv6 dhcp guard policy CLIENT

device-role client

!

vlan configuration 101

ipv6 dhcp guard attach-policy CLIENT

ACL: ipv6 access-list CLIENT_PORT

remark Block DHCPv6 Server on Client Ports

deny udp any eq 547 any

permit ipv6 any any

!

interface GigabitEthernet0/1

description Host Port

ipv6 traffic-filter HOST_PORT in

ROGUE DHCPV6 MITIGATION FIRST TRY


Does DHCPv6 Guard or an IPv6 ACL work?

Yes for non-malicious/non-fragmented DHCPv6 packets

As with RAs, DHCPv6 Guard and basic ACLs can be bypassed with the fragmentation evasion

But no known attack tools in the wild that have the fragmentation evasion built in

Howeverscapy could be used to craft an attack, but would be some work

ROGUE DHCPV6 MITIGATION EFFICACY


Mitigation options against fragmented DHCPv6 replies:

Option A block fragment evasion packets (initial packet) with undetermined-transport option

Option B block fragment evasion packets (non-initial packets) with crafted ACL

Unlike with RAs, DHCPv6 replies are unicast easier to block

DHCPv6 packets use a link-local address, so block fragments from:

- fe80::/64 (all link-local addresses)

Unlikely, only configure if in use

- fe80::/10 (defined link-local only fe80::/64 should be used but some systems allow)

ROGUE DHCPV6 MITIGATION SECOND TRY


Mitigation options against fragmented DHCPv6 replies:

Option A:

Option B:

ROGUE DHCPV6 MITIGATION SECOND TRY


An IPv6 subnet has over 18 quintillion addresses

Try to use up all the leases is futile, attackers wont wait for years

However, as with IPv4 you can limit the number of addresses leased per port with IPv6 snooping.

DHCPV6 FLOODING?


VPN Bypass


DHCPv6 Spoofing





ROADMAP


Attacker (aggressively) scans a network (e.g. scan6)

Scan triggers neighbor discovery (resolve address to MAC)

The theory is that the flood of NDP packets overwhelms the router/switch and thus a DoS

Issue?

Local scan (attacker scanning same VLAN) could be:

Not an IPv6 issue excessive L2 broadcast/multicast traffic can overload some switches

Solution throttle broadcast/multicast traffic on host ports to reasonable levels, e.g.:

- storm-control broadcast level 2.00 1.00

- storm-control multicast level 5.00 1.00

Also Destination Guard (next topic!)

REMOTE SCANNING DOS


Issue?

Remote scan:

Does not generally appear to be an issue for Cisco devices

IOS has a built in rate limiter (not tunable)

show ipv6 traffic - look under ICMP statistics, Sent, # output, # rate-limited

IOS limits incomplete NDP entries

show ipv6 neighbors statistics INCMP appears to be capped at 512

IOS appears to have separate caches for incomplete versus completed entries

Newer versions of IOS allow NDP cache tuning if desired:

ipv6 nd cache interface-limit #

REMOTE SCANNING DOS


But what if it could possibly be an issue?

Mitigations:

Ingress ACLs

Destination Guard the silver bullet!

Destination Guard:

Using IPv6 snooping/gleaning the switch learns all L2 neighbors

When an attacker tries to scan a subnet the switch can:

Summarily drop all requests for unknown neighbors

Only drop requests for unknown neighbors under stress

Cons? Only available on 4500s and 7600s today, but coming on other platforms.

REMOTE SCANNING DOS


VPN Bypass


DHCPv6 Spoofing/Flooding





ROADMAP


MONITORING AND CONTROLLING IPV6

Service Number Description IPv6 Encapsulation IPv4/41 Tunnel IPv6 over IPv4

Generic Tunnel IPv4/47 Tunnel anything over GRE

Teredo/Miredo UDP/3544 Tunnel IPv6 over UDP (NAT Traversal)

Teredo/Miredo Non-Standard IPv6 destination starting with 2001:0000::/32 over UDP over IPv4

TSP TCP|UDP/3653 IPv6 Tunnel Broker using the Tunnel Setup Protocol (RFC 5572)

AYIYA TCP|UDP/5072 IPv6 Tunnel Broker using Anything in Anything (www.sixxs.net/tools/ayiya/)

Public 6to4 Anycast Relay IPv4:192.88.99.1

Starting with IPv6 source address of 2002::/16 (6to4 is IPv6 over IPv4/41) Destined to 192.88.99.0/24 for IPv4

IPv6 Encapsulation TCP/443 IPv6 over IPv4 SSL Tunnel, many variants

IPv6 Ethertype 0x86DD Distinct from IPv4 Ethertype (0x0800)

DNS IPv6 Records Several AAAA, updated PTR records - can be transported over IPv4 or IPv6

Image source: gfi.com


VPN Bypass







ROADMAP


Block Tunneling IPv6 through IPv4 network:

If you dont want IPv6 traffic going through a firewall then explicitly block it!

EXAMPLE FIREWALL POLICY


Firewall Policy

Dont block all ICMPv6!!!

Simple Examples for transit traffic, can get more granular:

Reference NIST SP 800-119 (Section 3.5, Table 3-7)

Reference RFC 4890 (Recommendations for Filtering ICMPv6 Messages in Firewalls)

IPV6 ACCESS CONTROL


Router/Switch Policy

Dont block the NDPs NS/NA functionality or you will break IPv6!

ipv6 access-list Example1

permit any host 2001:db8::1

permit icmp any any nd-ns

permit icmp any any nd-na

deny ipv6 any any

IPV6 ACCESS CONTROL


VPN Bypass







ROADMAP


NAT Security Considerations:

Topology hiding

Pros Makes attacks more challenging

Cons Operational costs/complexity, impedes easy communication

Prevents inbound access without prior outbound access

Pros Protection against a poorly configured firewall/ACL

Cons Same as above

General Security Considerations:

Most security comes from stateful firewalls and application inspection

Most attacks/compromises are drive-bys or the result of user initiated activities which NAT offers no protection against

NAT SECURITY


WHY IPV6 AND NO NAT?

Address space

Should be a virtually unlimited supply think street addresses

Facilitates communication/collaboration

Innovation

NAT Gateways make innovation harder (mainly driven by insufficient address space)

Productivity (easy communication/collaboration) is a key business objective which NAT impedes


PROBLEMS WITH NAT

Some protocols do not work correctly through NAT and require fix-ups (ALGs) or extra configuration E.g. ICMP, FTP, SIP, H.323, RTSP, some VPNs

NAT breaks end-to-end connectivity Connection establishment and/or packet data requires a 3rd

party

Affects Voice Calls, Video Conferencing, file sharing, Collaboration, etc. For example, Skype, Facetime, Webex, and Microsoft Sharepoint Workspace work better without NAT.

Note: Multiple NAT tiers can totally break these applications

NAT for address overlap is technically challenging

Limits innovation, increases costs/barriers for new ideas/solutions


BENEFITS OF NAT

NAT simplifies changing ISPs (If PI Addresses not used)

NAT hides the network topology and foils many simple network scans NAT alone is not secure, but it has been a helpful safety net

against sloppy firewall policies

Without NAT, firewall policies must be more robust and actively managed

NAT can easily solve some complex network issues Multi-homing ISPs, return path selection, asymmetric routing

NAT is ubiquitous Today, software is developed with an expectation of NAT

Tomorrow?


THE HIDDEN COSTS OF NAT

Something to consider when evaluating NAT:


QUESTIONS

?

CDW Solutions Blog: cdwsolutionsblog.com

CDW Advanced Technology Services

cdw.com/services


Appendix

BACKUP SLIDES


Common IPv6 L2 Security Issues and Options:

IPV6 SECURITY COUNTERMEASURES

Issue Solution

Spoofed/Illegitimate RAs RA Guard (or PACL)

Spoofed NDP NA MLD Snooping, DHCPv6 Snooping, NDP Inspection, SeND

(Spoofed) Local NDP NS Flood

NDP Inspection, NDP Cache Limits, CoPP

(Spoofed) Remote NDP NS Flood

Ingress ACL, CoPP, NDP Cache Limits

(Spoofed) DAD Attack MLD Snooping, NDP Inspection

(Spoofed) DHCPv6 Attack DHCPv6 Guard

Spoofed/Illegitimate DHCPv6 Replies

DHCPv6 Guard (or PACL)


REDMONDS STANCE

Per the Microsoft IPv6 FAQ:

From Microsoft's perspective, IPv6 is a mandatory part of the Windows operating system and it is enabled and included in standard Windows service and application testing during the operating system development process. Because Windows was designed specifically with IPv6 present, Microsoft does not perform any testing to determine the effects of disabling IPv6. If IPv6 is disabled on Windows 7, Windows Vista, Windows Server 2008 R2, or Windows Server 2008, or later versions, some components will not function. Moreover, applications that you might not think are using IPv6such as Remote Assistance, HomeGroup, DirectAccess, and Windows Mailcould be.


DISABLING IPV6 IN WINDOWS

What breaks if IPv6 is disabled on Windows Vista and Later?

Hyper-V Cluster - It is not possible to add a new node to an existing cluster

TMG Server - RRAS breaks

Exchange - Mail flow & Installation problems

SBS Server - Exchange services fail to start & network shows offline

DirectAccess - Does not work

HomeGroup - Does not work

Applications using Windows Peer-to-Peer Networking will not work

5 IPv6 Attacks and Countermeasures v1.2

Documents

ipv6 traffic

organization vpn bypass

vpn session

remote user vpn client

ipv6 router advertisement

remote user vpn access

malicious vpn bypass

tunnel vpn client traffic